Advanced Image Processing

Figure. 3D surface measured by conventional two-wavelength holography (left) and Angstrom Vision method (right). The tallest peak is about 20 microns high.

The main problem of quantitative holographic measurements is the speckle noise. This noise had been studied for decades, and very efficient methods had been developed recently for suppressing it in amplitude signals.

 For holographic 3D shape measurement, the phase signal is of main interest;  we therefore developed efficient speckle noise suppressing algorithms for phase signals as well.

 The effect of speckle-noise suppression is illustrated in the following figure. It shows 3D surface screening results by conventional holographic two-wave method (left) and Angstrom Vision approach (right). The conventional method demonstrates heavy speckle-noise that limits 3D measurement accuracy.

 With the innovative Angstrom Vision method, the 3D shape measurement accuracy is at sub-micron level.

 It is also worth noting that our advanced image processing algorithms are implemented in a computationally efficient (“fast”) way. In turn, when being run on a conventional GPU or FPGA, it achieves real-time processing at video frame rates for 5-pixel depth-maps.